Connect with the Product Hunt community

Hello Product Hunt!

It s been a loooong time since I posted here hehe.

We're excited to welcome AI for UI as Designeer's second Diamond Sponsor

Created by @emilkowalski , the course teaches designers and engineers how to use AI to build interfaces that people actually want to use.
Checkout : aiforui.dev/?utm_source=designeer&utm_medium=sponsored&utm_campaign=ai-for-ui

We're super excited to welcome @Motionsites as Designeer's third Diamond Sponsor
Created by @viktor_oddy , MotionSites offers premium AI prompts and design resources for creating cinematic 3D and scroll-animated landing pages.

Checkout : motionsites.ai/?utm_source=designeer&utm_medium=sponsored&utm_campaign=motionsites

I'm three weeks from launch and I have a list of fourteen things I could still build. I know I can't build all of them. What I don't have is a rule for deciding.

The rule I've been using "cut anything that isn't core" is useless, because everything feels core when you're the one who built it.

What I'm trying to get to is a single non-negotiable: the one thing that, if it's missing, the product isn't the product. Everything else is a candidate for cutting.

So for the solo makers here: what was your one non-negotiable thing at launch, and what did you cut that you thought you couldn't? And looking back, did cutting it hurt, or did nobody notice?

Opposite of me being someone who translates, I interact with this particular AI in Japanese and everything that is kept also goes in Japanese, i.e., my major rules file with more than a hundred incident notes explaining the basis or reasoning of each rule in general, also, memory notes. I reserve English only for the things that are to be English, such as a forum post or an article in English written by me. It is funny thing though that, with this AI, it tends to speak English quite often with me after long sessions that I was immersed in English documentation. In order to stop it from happening, now, for instance, I set up a trigger where every time this AI answers me in a language which is not Japanese, it will rewrite the English part.

There are two lines on OpenAI s GPT-5.6 model page that are easy to miss if you call the model by its alias.

The first: the plain gpt-5.6 alias routes to GPT-5.6 Sol. If your code asks for gpt-5.6 because it looked like the sensible default, you re paying Sol rates: $4 per million input tokens, $20 per million output.

The second: Sol s pricing is labelled promotional, available at least through November 21, 2026 . The page doesn t say what it becomes after that.

Take an ordinary feature: 100,000 requests a month at 1,200 tokens in and 300 out. That s $480 of input plus $600 of output, so $1,080 a month, on a price that s only promised until the 21st.

Three weeks after launch, Mossy's little plant lives on about 20 desks every weekday and has offered over 2,400 breaks.

My wife and I build it after our day jobs, so the only way it spreads is people like you. If you know someone who sits all day and forgets to stand up, send them heymossy.com. One coworker is enough.

Lately, it feels like even the smallest project comes with an unwritten rule that you need 15 dependencies, Docker containers, Tailwind plugins, and a complex deployment pipeline before you even render a <div>.

My recent experience was when I wanted to build a simple 3D browser runner game over the weekend. I started with the "recommended" modern web setup and immediately got buried in conflicting bundler configs, SSR hydration errors on canvas components, and package bloat. So I ended up tearing half of it out, switching to a lightweight editor (Zed), drafting the math directly in AI Studio, and keeping the game loop isolated in a barebones Vite canvas before wrapping it in Next.js just for hosting. The moment I stripped away the modern boilerplate, the project actually moved fast.

What s a part of modern web dev or startup tech stacks that you think has become completely over-engineered for 90% of use cases?

DASCA continues to grow with this new update. With Version 2, we rebuilt the entire UI system. We also introduced Node Mode as a new workflow, our first step into a node-based system powered by WebGPU. This update ships the first beta release of Node. Across the board, we revisited effects and polished them. Your feedback matters to us.

octoscope 0.36.0 is out and this time the terminal has someone in it.

We just shipped Second Brain 3.4.0, and since the last Product Hunt update was 3.0, I wanted to share what has changed.

The biggest addition in 3.4 is Projects.

If you use AI across different products, clients, codebases, or areas of your life, one giant pool of memory can get messy fast. Projects let you keep related memories together, so your AI can pull from the right context without you having to explain everything again.

Hi everyone, maker of NM Signals here , thanks for following along!

I m building Signals to help you check what AI can access and read on your website, understand what needs fixing, and review how those same pages work for human visitors.

I d love to hear your experience: what (if anything) surprised you most when you tried the tool?

We re getting close to the final stretch of our Product Hunt launch, and we re currently fighting for a higher position.

If you haven t checked out @SaleSmartly yet, we d really appreciate it if you could take a look and support us with an upvote and a comment.

Every bit of support helps us move up and reach more people today.

https://www.producthunt.com/prod...

Three things came out of the launch comments, and all three are live.

The town says why. When a check finishes, up to 100 of the people it reached get one more question. Those who scrolled past or got annoyed pick a reason, and those who liked it say what stopped them. For a post, those who stopped also say what they would comment. The page splits the reasons into ones about the text and ones about the audience, so you can tell a weak opening from the wrong readers. Jev also answers a few yes or no questions about the text itself:

• Is there a concrete number, name or example?

• Is it clear what the reader should do?

• For a listing, does the first sentence say what is for sale?

Your own audience, in words. Describe whom the text is for, and only the residents who fit read it. "People who work in IT and are into startups" is 238 people, "people over 60" is 1,398. The town has no data on cities or company stages, so words about those do not count. The Audience tab shows how Jev read the description.

Go ask it for the best tool in your category. If you're not in the answer the reason is likely not your product.

Surfer put out a study in September. They went through 289,105 pages that AI answers cite. The more of those pages mention a brand the higher it sits in the answer. ChatGPT showed this the strongest. And the mentions that count most are in regular blog posts even though blogs are under a third of what gets cited.

We ve just launched the AnySearch Open Source Bounty Program.

We re opening CNY 100,000 in bounties for individual developers to help bring AnySearch to more open-source AI agent projects. This round includes 30 selected projects. For most of them, integration will happen through a Pull Request, while the exact process depends on each project s architecture and contribution guidelines.

Hi everyone, GhostDeck 1.1 just landed on the Mac App Store, and it is a free update for everyone who has the app.

What's new

- FLAC tags and covers. FLAC files now show their artist, title, album and embedded cover art, even when you open single files instead of a folder.

Netfox 0.27.0 is out

This one is about names

every app that uses OAuth shows you a consent screen.

here are the permissions this app is requesting. the list is there, the allow button is there, and almost nobody reads it before clicking.

We ve gotten pretty good at measuring what an LLM costs.
Tokens in.Tokens out.API bill.
But an agent does a lot more than generate tokens.
It might call three tools, run a search, query a database, invoke another agent, retry a failed step, and finally produce something useful.
So what did that piece of work actually cost?
That s the problem I ve been thinking about with Kopai.
Every agent run should have a ledger:
What did it do?
What did it consume?
What did it cost?
And eventually what did it earn?
Not just token accounting.
A financial ledger for agentic work.
Tell me what you are using to monetise you Agents

Netfox 0.28.0 is out - https://youtu.be/_E3OsO35dbI?si=...

This one is about telling devices apart

Since July, App Store Connect asks whether your app has social media capabilities. From September, you can't submit a new app or an update without answering.

Apple posted on 9 July that the age rating questionnaire now includes questions about social media capabilities, and that "beginning in September 2026, responses will be required when submitting new apps or updates to the App Store, or when submitting apps for notarization for alternative distribution." September ends tomorrow, so if you have a build in flight, this is the week you meet it.

DungeonQ's defensive deception runtime keeps designated sessions working inside persistent synthetic worlds. What happens when that path fails matters as much as the successful walkthrough.

In the self-hosted reference, a synthetic transport failure does not fall back to the protected origin. Invalid authority, missing evidence, or unreadable state return bounded errors. The gateway records an attempt before dispatch, so an interrupted operation cannot simply disappear from the evidence.

The published September 18 acceptance record includes a useful negative check: with the containers stopped, verification returned INCONCLUSIVE instead of reusing an earlier passing result.

For operators, preserve the failed run's records and inspect the uncertainty. Automatic reconciliation of unknown outcomes is not implemented; deleting state is not a recovery procedure.

Everyone talks about how fast AI can write code now but nobody talks about the mental exhaustion of reviewing three different feature branches at the same time.

I realized last week that if I run more than two agents simultaneously I stop catching small architectural mistakes and just start blindly approving diffs.

Now I cap myself strictly based on how many code reviews I can realistically handle before my focus drops. Do you use any specific workflow or rule to keep yourself from drowning in unreviewed AI-generated code?

We re live today @SaleSmartly

SaleSmartly helps businesses manage customer conversations across channels like WhatsApp, Instagram, Messenger, and Telegram from one place, with AI agents and automation to streamline support and sales.

If you have a moment, please support us with an upvote and comment. Also, check out the product and share your feedback.